RORB Data: pseudo-centroids from annotation polygons using Gaussian weighted average over 11x11x11 cubes ≈ 0.61 microns^3 followed by step-wise regression.
2018-02-28
## Start: AIC=-2358.71
## gaba ~ MBP + TdTomato + DAPI1 + DAPI2 + DAPI3 + GluN1 + PSD95 +
## synapsin + VGlut1 + GABA + GAD2 + Gephyrin
##
## Df Sum of Sq RSS AIC
## - DAPI2 1 0.0230 39.221 -2360.2
## - DAPI3 1 0.0361 39.234 -2360.0
## - MBP 1 0.0392 39.237 -2359.9
## - DAPI1 1 0.0404 39.238 -2359.9
## <none> 39.198 -2358.7
## - GluN1 1 0.1676 39.365 -2357.3
## - TdTomato 1 0.3268 39.525 -2354.1
## - VGlut1 1 0.4278 39.626 -2352.1
## - GAD2 1 0.6291 39.827 -2348.1
## - synapsin 1 1.4834 40.681 -2331.2
## - PSD95 1 1.7254 40.923 -2326.6
## - Gephyrin 1 1.9041 41.102 -2323.1
## - GABA 1 7.7406 46.938 -2217.8
##
## Step: AIC=-2360.25
## gaba ~ MBP + TdTomato + DAPI1 + DAPI3 + GluN1 + PSD95 + synapsin +
## VGlut1 + GABA + GAD2 + Gephyrin
##
## Df Sum of Sq RSS AIC
## - DAPI3 1 0.0312 39.252 -2361.6
## - DAPI1 1 0.0350 39.256 -2361.5
## - MBP 1 0.0366 39.257 -2361.5
## <none> 39.221 -2360.2
## - GluN1 1 0.2800 39.501 -2356.6
## - TdTomato 1 0.3547 39.576 -2355.1
## - VGlut1 1 0.4583 39.679 -2353.0
## - GAD2 1 0.6159 39.837 -2349.9
## - synapsin 1 1.5572 40.778 -2331.4
## - PSD95 1 1.7025 40.923 -2328.6
## - Gephyrin 1 1.9098 41.131 -2324.5
## - GABA 1 8.2289 47.450 -2211.2
##
## Step: AIC=-2361.62
## gaba ~ MBP + TdTomato + DAPI1 + GluN1 + PSD95 + synapsin + VGlut1 +
## GABA + GAD2 + Gephyrin
##
## Df Sum of Sq RSS AIC
## - MBP 1 0.0132 39.265 -2363.3
## - DAPI1 1 0.0416 39.294 -2362.8
## <none> 39.252 -2361.6
## - GluN1 1 0.2560 39.508 -2358.5
## - TdTomato 1 0.3696 39.622 -2356.2
## - VGlut1 1 0.4596 39.712 -2354.4
## - GAD2 1 0.5954 39.847 -2351.7
## - synapsin 1 1.5831 40.835 -2332.3
## - PSD95 1 1.6744 40.926 -2330.5
## - Gephyrin 1 1.9575 41.209 -2325.0
## - GABA 1 10.4930 49.745 -2175.8
##
## Step: AIC=-2363.35
## gaba ~ TdTomato + DAPI1 + GluN1 + PSD95 + synapsin + VGlut1 +
## GABA + GAD2 + Gephyrin
##
## Df Sum of Sq RSS AIC
## - DAPI1 1 0.0420 39.307 -2364.5
## <none> 39.265 -2363.3
## - GluN1 1 0.2625 39.528 -2360.1
## - TdTomato 1 0.3690 39.634 -2357.9
## - VGlut1 1 0.4539 39.719 -2356.2
## - GAD2 1 0.6124 39.878 -2353.1
## - synapsin 1 1.5796 40.845 -2334.1
## - PSD95 1 1.6655 40.931 -2332.4
## - Gephyrin 1 1.9648 41.230 -2326.6
## - GABA 1 10.4806 49.746 -2177.7
##
## Step: AIC=-2364.5
## gaba ~ TdTomato + GluN1 + PSD95 + synapsin + VGlut1 + GABA +
## GAD2 + Gephyrin
##
## Df Sum of Sq RSS AIC
## <none> 39.307 -2364.5
## - GluN1 1 0.3166 39.624 -2360.1
## - TdTomato 1 0.3615 39.669 -2359.2
## - VGlut1 1 0.5043 39.812 -2356.4
## - GAD2 1 0.6551 39.962 -2353.4
## - synapsin 1 1.5402 40.847 -2336.0
## - PSD95 1 1.7363 41.044 -2332.2
## - Gephyrin 1 2.1129 41.420 -2325.0
## - GABA 1 10.4904 49.798 -2178.9##
## Call:
## lm(formula = gaba ~ TdTomato + GluN1 + PSD95 + synapsin + VGlut1 +
## GABA + GAD2 + Gephyrin, data = sdat)
##
## Residuals:
## Min 1Q Median 3Q Max
## -1.53487 -0.08879 -0.04081 0.02644 0.93385
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 0.107188 0.007951 13.480 < 2e-16 ***
## TdTomato -0.022543 0.008395 -2.685 0.00740 **
## GluN1 -0.024721 0.009837 -2.513 0.01217 *
## PSD95 -0.055816 0.009485 -5.885 5.90e-09 ***
## synapsin -0.051453 0.009283 -5.543 4.07e-08 ***
## VGlut1 -0.031204 0.009838 -3.172 0.00157 **
## GABA 0.162002 0.011200 14.465 < 2e-16 ***
## GAD2 0.040866 0.011305 3.615 0.00032 ***
## Gephyrin 0.056079 0.008639 6.492 1.50e-10 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 0.2239 on 784 degrees of freedom
## Multiple R-squared: 0.482, Adjusted R-squared: 0.4768
## F-statistic: 91.2 on 8 and 784 DF, p-value: < 2.2e-161 Results
1.1 1-d Heatmap
1.2 Location meda_plots
1.3 Outliers as given by randomForest
1.4 Correlation Matrix
1.5 Cumulative Variance with Elbows
1.6 Paired Hex-binned plot
1.7 Hierarchical GMM Classifications
1.8 Hierarchical GMM Dendrogram
1.9 Stacked Means
1.10 Cluster Means
2 Restricting hGMM to \(K = 2\)
Here we are restricting hierarchical GMM to only go through on level. We are comparing the cluster results to the gaba labels.
set.seed(3144)
h2 <- hmc(sdat[, -1], maxDepth = 2, ccol = ccol, model = c("VVV"))
h2lab <- viridis(max(h2$dat$labels$col))
h2col <- h2$dat$labels$col2.1 K = 2 stacked means plot
p1 <- stackM(h2, ccol = ccol, centered = TRUE, depth = 1)
p1
2.2 Pairs plot colored by true gaba classification
cols <- c("black", "magenta")[gabaID$gaba+1]
acols <- alpha(cols, 0.35)
#pairs(h2$dat$data, pch = 19, cex = 0.7, col = acols)
#plot(h2$dat$data, col = acols, pch = c(19,3)[gaba+1], cex = c(0.5,1)[gaba+1])
pairs(sdat[,-1], col = acols, pch = c(19,3)[gaba+1], cex = c(0.5,1)[gaba+1])
2.3 Pairs plot colored by hGMM cluster classification
acols2 <- alpha(h2lab[h2$dat$labels$col], 0.45)
par(bg = "gray45")
#plot(h2$dat$data, pch = c(3,20)[gaba + 1], cex = 1, col = acols2)
pairs(sdat[,-1], pch = 19, cex = 0.7, col = acols2)
dev.off()## null device
## 13 Permutation test for ARI
p0 <- mclust::adjustedRandIndex(pred, gaba)
perms <- foreach(i = 1:1.5e4, .combine = c) %dopar% {
set.seed(i*2)
mclust::adjustedRandIndex(sample(pred), gaba)
}
pv0 <- sum(c(perms,p0) >= p0)/length(perms)hist(perms, xlim = c(min(perms), p0 + 0.25*p0),
main = "permutation test of ARI values", probability = TRUE)
#hist(perms, probability = TRUE)
abline(v = p0, col = 'red')
t1## truth
## pred FALSE TRUE
## FALSE 590 9
## TRUE 118 764 Summary Table
| measurment | value |
|---|---|
| Misclassification Rate | 0.1601513 |
| Accuracy | 0.8398487 |
| Sensitivity | 0.8941176 |
| Specificity | 0.8333333 |
| Precision | 0.3917526 |
| Recall | 0.8941176 |
| ARI | 0.3584828 |
| \(p\)-value for ARI | 0.000067 |
| F1-score | 0.5448029 |
| TP | 76 |
| FP | 118 |
| TN | 590 |
| FN | 9 |